- Количество слайдов: 9
draft-weil-shared-transition -space-request-00 Jason Weil, Victor Kuarsingh, Chris Donley, Christopher Liljenstolpe, Marla Azinger
• • • Underlying Drivers In the IPv 4 to IPv 6 transition period, end-user SPs will have to support heritage IPv 4 customer end devices (think slingboxes, game consoles, SNMP managed toasters, etc. ) Contrary to popular belief, operators don’t like blowing up customers’ services on purpose - we do it well enough by accident, therefore we need to support those devices IVI, DSlite, etc. could support those heritage devices, but require CPE / RG changes - and those devices are just now starting to arrive in the market.
But it’s NAT. . • For many of the carriers represented here (and others), NAT 444 is the least disruptive way of managing those heritage devices during the transition. • Yea, we know, trust us, if there was another path, we wouldn’t be inflicting this pain on ourselves. . • As soon as possible, we want that NAT gone as well, but millions (or tens of millions) of CPE fleet changes takes time.
What do we need? • A re-useable address block (/10 is proposed) to allow the CPE to CGN network to be deconflicted with the customers’ networks. • Best place to assign from - IANA. • Can be discontiguous, but would prefer not.
Use and Constraints • Used for any transition mechanism • Limited to use by infrastructure providers - we know other users will poach it, • but people that do will be “off the reservation” • no default or example configs, no end-user documentation mentioning it, etc. will limit that poaching to “knowledgeable” users.
Why longer than a /8 • Pragmatism • Limit the impact on the remaining pool • Some operators need multiple /8’s, some need one /8, some less. That means that some operators will have to have pools of NAT. • Yes, we know it’s painful, refer to slide 3
But it’s still NAT. . . • Yes, it will still be broken, and will still suck • But it will suck less, and base capabilities of heritage v 4 elements will still work.
Constructive suggestions by the community • Break our networks and services to force customers to Thank’s a paycheck upgrade to IPv 6 - sorry, we like to get • • Consume all remaining IPv 4 address space to force the run -out crisis - sorry, don’t want to go down as “the folks who completely borked Geoff’s predictions (and the Net)” Deploy (or force our customers to buy) 10’s of millions of barely available (or unavailable) CPE in the next six months - refer to point one. We’d probably singularly support the largest build-out in outsourced technical support call centres, however.
But it’s NAT. . . • We know, remember • Get over it • We will want it gone more than you - support will be be corrosive, but support headaches are better than not needing any support what so ever.